Talkative Chatbot Builder

Created by Seb Coulthread, Modified on Tue, 23 Apr 2024 at 09:52 PM by Seb Coulthread

How to build chatbots in Talkative!

This is a guide for how to use the Talkative chatbot builder. This is available on all Talkative subscriptions.

Talkative's chatbot builder is only accessible by Account Holders.

Creating a Chatbot

First, navigate to Settings and search for "chatbot":

Create a chatbot:

Queue emails are used if the chatbot is sending a form to an email address.

A timeout is not needed but is recommended at 10 minutes. This means chats will end after that amount of inactivity. If a chatbot is transferred or queued, this time limit no longer applies.

You can edit these settings at any time, or create more chatbots.

Once your chatbot has been created, it will show up in your list of queues and users:

Before you go live with your chatbot, you will need to map it to your Chat Widget, which is done in your Chat Widget editor.

Building Chatbots - Key Concepts

The chatbot is built up with a number of "Nodes". Generally, each Node consists of
• Actions, i.e. what the chatbot is doing on that node and;

• an Exit Approach, i.e. how you get to the next node.

Each Node has a label, labels are used purely for building purposes and are not exposed to customers.

Actions

Each node allows you to set any number of the following actions:

•Send Message(s) - a text-based message that is sent to the customer. You can include Interaction Data using the syntax where you replace dataName with the name of the data you want to include. In the example below we can provide a personalised greeting to the customer if we already know the fullname variable.

• Send File - a file that is sent to the customer. You can upload any file type, there is a 10MB size limit.

• Make a web request and write int. data: https://support.gettalkative.com/knowledge/talkative-chatbot-fulfillment-webhooks

• Custom Message Suggestions - if you want to provide custom message suggestions.

• Write Interaction Data - if you want to take the user's input and have it create interaction data, for example collecting an email address. Interaction data can be set to Static Text (a fixed value), Entire Previous Message, Email in Previous Message (the system auto-detects email values), or custom regex in previous message (useful for detecting things like order numbers or account numbers).

Exit Approaches

Exit approaches are how the system decides which node to go to next. There are multiple options:

• Redirect to next node immediately - do not wait for the user input
• Branch off based on user input - the user's message defines which node follows
• Always go to next item - wait for the user to message, then go to the next input
• Transfer to queue - Transfers will transfer regardless of queue presence. So it's recommended that you check the queue status first in your chatbot flow.
• Route based on interaction data value
• Route based on interaction data existence - for example only transfer to the sales queue if you have first captured the user's email address
• Based on queue status - useful if you want to vary the message flow depending on whether your queue is online (or not)
• Redirect to node
• Lookup previous message in AI Knowledgebase: https://support.gettalkative.com/knowledge/talkative-genai-chatbot
• Escalate to email (web only)
• End interaction
• Redirect to the start of the chatbot flow

Warning: changing the exit approach will delete all descendents of this node. If you need to change an exit approach but want to retain the chatbot nodes below, we'd recommend using copy/paste on the nodes you want to preserve.

Training Phrases

When selecting "Branch off based on user input", the chatbot works by matching the user's text input to Training Phrases within the target node. The user's input can be from free-text input or from custom suggestions (more info on this below).

You can have multiple training phrases - the more you have, the more effective your chatbot will be at matching user input with intents/nodes.

AI matching to Training Phrases

The chatbot uses AI to improve the matching of customer text input to nodes.

It normalises the customer message and the training phrases, which means capitalised words and words with extra punctuation are matched.

If an exact match isn't found after normalisation, AI checks if the intention of the customer message is close to any of the phrases.

If you do not want messages being sent to an external AI service for this, please disable the AI enhancement config (set as part of the chatbot create/edit form).

Message Suggestions

If you select the Exit Approach of "Branch off based on user input" and you enter in the node names, these will show up as message suggestions in the original node.

In the example below, 3 nodes have been created, and their titles (Sales, Support, Speak to the Team) are automatically added as suggestions in the Welcome node:

You can add suggestions, which will create a new node below:

Using the standard message suggestions, you cannot delete suggestions. If you want to have custom suggestions, or only display certain suggestions to users, click "Add Action" and then click "Custom Message Suggestions".

Fallbacks / Retries

Fallbacks are defined as the node that is hit if the chatbot does not understand the user's input.

The Fallback node is set at the top of the tree on your first node.

You can have multiple retry messages.

Maximum retry attempts counts the number of consecutive retries, rather than aggregate. For example, if you had retries set to 3, and then entered your email address incorrectly 2 times, but then entered it successfully on the 3rd attempt, then got to another fallback, the retry counter would be re-set to 0.

Errors

If your chatbot has any errors in the logic/set up, you will see a warning in the top right corner.

Click on this to take you to the issue node:

Previews

At any time you can preview the chatbot to test it out as a user. In order to use preview you will need to create a widget config.

Saving, Publishing, & Versions

You should regularly save changes as you edit your chatbot.

You should only publish after checking the chatbot flows. Once you publish, your changes will be live for all of your users!

Shortcuts

Talkative supports several keyboard shortcuts:

Ctrl Z - Undo

Ctrl Y - Redo

Ctrl C - Copy

Ctrl V - Paste

Ctrl M - Minimise nodes

Ctrl P - Search

Ctrl S - Save

Ctrl Shift S - Publish

Ctrl+ Enter - Preview from start

Ctrl+Shift+Enter - Preview from current node

Ctrl B - Exit preview

Ctrl H - Scroll current node into view

Ctrl + - Zoom in

Ctrl - - Zoom out

Backspace/delete - Delete current node

Search is useful if you want to find a specific message or node:

Minimise can be useful to get a simplified view of your chatbot:

Managing Versions

From the top level menu, you can access historical chatbot versions, if you need to revert to a previous version, or duplicate a previous version for editing:

Roadmap

As of writing (Q4 2023), Talkative is currently developing a number of enhancements:

• AI-generated training phrases

• AI nodes - respond with AI-generated based on a knowledgebase/website

• Rephrase responses each time using generative AI

• AI matching of training phrases

• Generate example chatbot flows and other insights based on previous month's data

• Fulfilment - lookups into your own CRM/data

FAQs:

Q: How many chatbots can I have?

A: There are currently no limits. You can build multiple different chatbots to suit different use cases.

Q: How do I delete one node without deleting everything underneath?

A: The best way to do this is Copy everything underneath that you want to keep, delete the node you want to delete, then Paste.

Q: How do I delete some suggestion chips?

A: You will need to click "Add Action" and then click "Custom Message Suggestions".